OpenAI has deployed a new automated security testing system for ChatGPT Atlas, but has also conceded that prompt injection remains an "unsolved" security threat.
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback